In Xeqmate AI video surveillance, it is the server that analyses the footage — in the cloud or on-premise. That means any camera with an RTSP or RTMP stream gets person, vehicle, object, fire and PPE detection analytics, with no hardware swap. And the fine tuning — zones, trigger conditions, sensitivity and schedules — is what separates useful alerts from a flood of false alarms.
In AI video surveillance, algorithms analyse the camera video and raise events when they detect something relevant — a person, a vehicle, fire. At Xeqmate that processing runs on the server (in the cloud or on-premise), not on the camera: any camera sending video over RTSP or RTMP gets the analytics, with no special hardware. The operator stops watching dozens of screens and starts handling events.
That architecture changes the economics of the project: instead of buying "smart" cameras for every point, intelligence becomes a service applied to the cameras you already have. The same principle holds for the platform specialised AI — license plate recognition (LPR) and facial recognition can also be processed on the server from an ordinary stream.
Each camera can run up to 2 detection algorithms at once — pick the ones that actually make someone act.
Activity in the defined zones — the baseline analytic for scenes that should be still.
Human presence in the monitored zone — the classic after-hours intrusion detection.
Vehicles in areas of interest — yards, access points and restricted circulation zones.
Objects detected inside the zones drawn over the camera image.
Signs of fire in the image — an alert that cannot depend on somebody watching the screen.
Personal protective equipment in operational areas — AI applied to workplace safety.
False alarms are what kill a monitoring operation. Xeqmate tuning attacks the causes, camera by camera.
Draw the regions that matter over a real frame from the camera — each zone with a name and a colour. Everything outside them is ignored: that is how you exclude the busy street behind the gate or the tree swaying in the wind.
The event can require a minimum number of objects (2 or more people together, say), consecutive detections (which cuts triggers from a passing blur) or a time window with continuous presence (someone standing in the zone for 5 seconds).
Low, medium or high — the balance between catching and crying wolf. Our advice is to start at medium with tightly cut zones: most false positives come from an oversized zone, not from sensitivity.
Days of the week and time ranges when the analytic works. The classic: person detection only outside business hours — during the working day, people moving around is normal.
Search the captures by detection type, camera and period, with the markings drawn on the image — find the event without watching hours of recording.
Every incident is handled with a status change and a comment — the operation knows what was seen, by whom and what was done. Meet the Event Center.
Push, popup and Telegram (profile and groups) take the trigger to the right people, right away.
For anyone running many cameras across many customers, this flow is the heart of the service — see how the platform serves central stations and how the analytics fit into the complete VMS SaaS, next to recording, permissions and auditing.
Every vehicle becomes a searchable read: plate alerts, search by photo or voice, and convoy, co-occurrence, flow and heat-map reports.
Explore LPRReal-time face monitoring, face alerts, face comparison and a persons-of-interest module — with privacy governance built in.
Explore facialNo. Xeqmate analytics run on the server — in the cloud or on-premise. Any camera sending video over RTSP or RTMP gets motion, person, vehicle, object, fire and PPE detection. What lens, height and distance change is the quality of the scene, not compatibility.
Motion, person, vehicle, object, fire and PPE detection — up to 2 algorithms per camera. On top of those, the platform has specialised AI for license plate recognition and facial recognition.
Through per-camera tuning: up to 4 detection zones drawn over the image (everything outside is ignored), a trigger condition (minimum count, consecutive detections or a time window with continuous presence), sensitivity in three levels, and operating hours. In practice, most false positives come from an oversized detection zone — cutting the zone tightly fixes more than touching sensitivity.
Yes. Operating hours define the days of the week and the time ranges in which the analytic works. The classic use is person detection only outside business hours: during the day, people moving around is normal; at night, it is an event.
In three places: Smart Search (search by detection type, camera and period, with the markings on the image), the Event Center (handling incidents with a status and a comment) and the notifications — push, popup and Telegram.
Yes. Processing runs on a server in the cloud or inside the customer infrastructure — the same analytics, the same tuning and the same interface. Cloud and local servers live on the same platform.
30 minutes with our team: we connect one of your cameras live, switch on an analytic and you watch the events arrive — before deciding anything.
We reply within one business day · no credit card, no install, no commitment